Content authentication is a significant problem of the research of image security. This article exploits visual and structural feature maps to construct a new image hashing algorithm for content authentication. The visual feature map is extracted by the luminance contrast model. Since the visual feature map can reflect image regions of human visual attention, the features of the tampered areas are captured in the hash. In addition, the structural feature map is extracted by the dual-tree complex wavelet transform. Next, the block Krylov singular value decomposition is applied to the visual feature map and the structural feature map for constructing feature vectors. Finally, feature vector distances are calculated and encrypted to produce hash. Extensive experiments are performed on the public databases to demonstrate effectiveness. The results demonstrate that the proposed algorithm can correctly detect 89.83% similar images and 93.93% forged images under the optimal threshold. The comparative results show that the proposed algorithm outperforms some baseline algorithms in content authentication.

Hashing is an effective approach for processing images, but most existing schemes have limitation in recognizing images undergone compound distortions, especially the compound distortions with rotation. To address this, we propose a rotation-invariant image hashing scheme based on an adaptive weighted feature space. The proposed scheme extracts three complementary feature maps from the input image, including lightness, color, and geometric structure. Each map is assigned an adaptive weight according to its relative contribution, and the weighted maps are fused into a unified representation. To ensure rotation invariance, the fused image is partitioned into concentric circular regions that maintain spatial alignment under rotation. The Quaternion Generic Fourier Transform (QGFT) is then exploited to extract frequency-domain descriptors. Finally, the hash sequence is generated through quantization. Extensive experiments demonstrate that the proposed scheme achieves strong robustness and discriminability, with an AUC of 0.99851 under large-angle rotations and a PRAUC of 0.92619 in copy detection, outperforming several representative hashing schemes.

Screen content image quality assessment (SCIQA) is an important topic of computer vision. Most existing no-reference SCIQA (NR-SCIQA) methods cannot effectively capture features of the texts and graphics in screen content images (SCIs) and thus have limitations in accurate quality prediction. This article proposes a novel edge and visual salient feature fusion network for NR-SCIQA. An important contribution is the design of the edge feature extraction module (EFEM) and the semantic feature extraction module (SFEM), which can accurately capture the potential edge features and high-level semantic features in SCIs, thereby constructing a hierarchical feature representation that spans from low-level edge features to high-level semantic features. Another contribution is the visual feature extraction module (VFEM), which can extract textual information that is more consistent with human visual salient features from SCIs. Experiments show that our proposed method performs well on two publicly available SCI datasets. Comparisons demonstrate that our proposed method surpasses some existing NR-SCIQA methods in quality prediction.

No-Reference Image Quality Assessment (NR-IQA) is an important task in the field of computer vision. Most methods utilize the pre-trained features with information irrelevant to image quality. In addition, some methods directly regress the pre-trained features without interaction or simply concatenate all features for score regression. They ignore the differences between local and global features. These issues lead to the limited IQA performance. To address this, we propose a Hybrid-Attention Refinement and Hierarchical Fusion Network (HAR-HFNet) for NR-IQA, which consists of a feature extraction module, a hybrid attention refinement module, a hierarchical dilated-attention fusion module and a quality prediction module. Firstly, the hybrid attention refinement module filters and refines the multi-stage features extracted by the pre-trained Swin Transformer, which enhances distortion-related information. Secondly, the hierarchical dilated-attention fusion module fuses the deep global feature with local features. It enables effective hierarchical integration of global semantics and local details. Finally, the quality prediction module predicts the score through weighted feature aggregation. Experiments on six public IQA datasets demonstrate that the HAR-HFNet outperforms some baseline NR-IQA methods in prediction accuracy and generalization ability.

Copy detection is crucial for image copyright protection. This paper proposes an effective image hashing approach via Saliency Model and Tucker Decomposition (SMTD) for copy detection. The first contribution is the third-order tensor construction with the saliency model called the Luminance Contrast (LC) model. As the LC model is capable of detecting image salient regions, the third-order tensor based on this model can enhance the robustness of our SMTD approach. Another contribution is the hash learning from the third-order tensor via the tensor decomposition called Tucker Decomposition (TD). Since the TD can retain the topological information of the original tensor, the intermediate hash learned from the TD result is discriminative. In addition, the intermediate hash is encrypted and quantified by segmented linear iterative chaotic mapping to generate the final secure hash. Numerous experiments are performed to confirm the effectiveness and superiority of our SMTD approach. The results demonstrate that our SMTD approach achieves better classification and copy detection performances compared with several existing benchmark approaches. Complexity analysis shows that our SMTD approach has low computational and storage costs.

Fine-Grained Image Retrieval (FGIR) is an important visual task. Most existing hash-based FGIR methods have difficulty in sufficiently leveraging low-level attributes and effectively associating high-level semantic information with these low-level attributes, thereby limiting their ability to accurately discriminate fine-grained images. To address this issue, we propose a Semantic-Guided Multi-Representation Fusion Hashing (hereafter SMRFHash), which uses semantic features as an anchor to guide and align color and frequency features, thereby enhancing fine-grained retrieval. Specifically, a multi-branch feature extraction module is proposed to separately capture frequency, color, and semantic features of images by attention mechanisms. Then, we further develop a semantic-guided attention fusion module that guides the model to focus on frequency and color regions relevant to the target object. In addition, we propose a confidence-weighted hash loss, which jointly exploits the triplet loss and cross-entropy loss to constrain network training. The proposed loss dynamically re-weights training samples according to the model’s prediction confidence. It improves early-stage training stability and enhances discriminative learning from hard samples. Experimental results show that the SMRFHash outperforms several state-of-the-art methods in retrieval performance.

Screen Content Image Quality Assessment (SCIQA) plays a significant role in computer vision. Current No-Reference SCIQA (NR-SCIQA) methods often struggle to effectively represent both textual and pictorial regions in Screen Content Images (SCIs), which hinders accurate quality prediction. This paper presents a NR-SCIQA method based on regional distortion perception. A key contribution is that we propose two feature perception modules: Textual Feature Perception Module (TFPM) and Pictorial Feature Perception Module (PFPM). The TFPM uses Discrete Wavelet Transform (DWT) and Laplace operators to refine structural and edge information, and the PFPM combines depthwise separable convolution with Discrete Cosine Transform (DCT) for joint spatial-frequency feature extraction. Another contribution is the Text-Picture Feature Fusion Module (TPFFM) which incorporates attention and adaptive mechanisms to integrate perceived features from both regions. Experimental results on the SIQAD and SCID datasets show that our proposed method significantly outperforms existing NR-SCIQA methods.

Copy detection is an important task of video copyright protection. This paper proposes a robust video hashing with discrete wavelet transform (DWT) and tensor Singular Value Decomposition (SVD) for copy detection. In the proposed algorithm, the video frames are evenly grouped and the frames in each group are divided into the red, green and blue channels. The means of these three channels are calculated separately, and robust secondary frames are obtained after polar coordinate transformation. Next, the robust secondary frames are input into the Vision Transformer to obtain feature maps and the tensors are constructed for extracting the deep features via tensor SVD. In addition, the robust secondary frames of all frame groups are integrated into a tensor, and a low-frequency subband is obtained by using three-dimensional DWT for extracting the structural features. Finally, the structural and deep features are binarized and concatenated to obtain a hash. Comparative experiments of copy detection and classification verify that the proposed algorithm outperforms some baseline algorithms.

No-Reference (NR) Image Quality Assessment (IQA) is a key technology of computer vision that enables the direct evaluation of image quality in the absence of a reference image. Most existing NR-IQA methods use a single attention mechanism to integrate the convolutional neural network (CNN) and vision Transformer, which makes it difficult to leverage the advantages of different attention mechanisms in integrating multi-scale distorted features. To address this, we propose an NR-IQA network based on Feature-Mixed Enhancement and Multi-Attention Feature Fusion. The feature-mixed enhancement module consists of a Multi-scale Feature Enhancement Block and an Interactive Feature Enhancement Block. This module is proposed to refine multi-scale CNN features and promote feature interaction. It can obtain accurate distortion features for better IQA. The multi-attention feature fusion module is composed of a Hybrid Attention Fusion block and a Multi-scale Attention Fusion block. This module uses three types of attention to fuse global-local information from three perspectives. Experimental results demonstrate that the proposed method surpasses some baseline NR-IQA methods in both prediction accuracy and generalization ability.

No-Reference Image Quality Assessment (NR-IQA) serves as a fundamental task in computer vision that aims to predict image quality consistent with human perception. Currently, numerous NR-IQA methods often use simplistic fusion strategies to integrate features from different backbones. However, these methods typically neglect the semantic differences and intricate inter-channel interactions among different features, thereby limiting their abilities to represent features effectively. To address this issue, we propose a Semantic-guided Channel Cross-attention Integration Network for NR-IQA (SCCIN-IQA), which enables more effective integration of complementary information from different backbones. The core module of our method is the fusion-semantic channel cross-attention. It first generates a semantic feature by integrating features from different backbones, then utilizes this semantic feature as a query to integrate the original backbone features via a channel cross-attention mechanism, thereby adaptively highlighting quality-relevant channel activations. Additionally, a space-channel enhancement module is introduced to further enhance the learned features in both space and channel dimensions, enabling comprehensive modeling of multi-dimensional contextual dependencies. Extensive experiments conducted on multiple public datasets demonstrate that the proposed SCCIN-IQA achieves state-of-the-art performance, consistently surpassing several mainstream methods while exhibiting strong generalization.

Self-Supervised Video Hashing (SSVH) has been widely applied to efficient video retrieval. Existing methods mostly use Siamese-like pairwise training to model inter-sample similarities, yet lack direct optimization for high-quality video content representation, limiting practical performance. To address this issue, we propose a novel SSVH method based on the Mamba-Transformer network (hereafter VHMT), which employs a teacher-student architecture. The teacher model extracts robust low-frequency semantic features using the Discrete Wavelet Transform (DWT) and captures temporal dependencies through a Vision Transformer (ViT)-based Temporal Attention Module (TAM). Concurrently, the introduced Temporal Reconstruction Enhancement (TRE) block further improves fine-grained reconstruction, thereby generating high-quality temporal modeling signals. The student model employs the Mamba architecture to model temporal relationships and achieves knowledge transfer and efficient single-path inference by approximating the teacher model’s feature reconstruction and attention outputs. Furthermore, we introduce a joint loss function to enhance the student model’s approximation of the teacher model’s temporal modeling capability while optimizing the clustering structure of the generated hash codes. Experimental results demonstrate the VHMT outperforms several state-of-the-art methods in mean Average Precision.

Video hashing is a helpful technique to address the copy detection issue. We propose a robust video hashing with visual feature maps and isometric mapping (Isomap) for copy detection. In the proposed algorithm, all frames of the video are converted to the HSI color space, and the intensity components are selected to represent the original frames. Next, the intensity components of all frames are divided into multiple groups, and the mean of each group is calculated on the third dimension to obtain grayscale secondary frames. Then, the luminance contrast model calculates visual feature maps from the discrete wavelet transform domain of the grayscale secondary frames, and a visual matrix based on the visual feature maps is constructed. The Isomap is used to learn low-dimensional features from the visual matrix. Finally, the low-dimensional features are quantified using ordinal measures. The experimental results on two public databases verify the classification performance of the proposed algorithm and its effectiveness in copy detection.

No-reference image quality assessment (NR-IQA) is an important task of computer vision. Most deep neural networks based NR-IQA methods have the ability of accurate quality predictions, but they have large-scale parameters and high computational complexity. To alleviate these problems, we propose a lightweight NR-IQA method by unifying statistical and refined semantic features. Our proposed method consists of a lightweight feature extractor (LFE), a statistical semantic feature extraction (SSFE) module, and a refined semantic feature extraction (RSFE) module. The LFE is used to extract semantic features with perceptual distortion information. The SSFE module is designed to obtain statistical information of the semantic features for capturing the local and global changes of distorted image. The RSFE module is designed to refine the semantic features for measuring complex distortions. Extensive experiments on many image quality assessment (IQA) datasets are done and the results indicate that our proposed method outperforms some baseline NR-IQA methods in IQA performance, generalization ability, and model complexity.

Most data-driven deep learning-based Artistic Image Aesthetics Assessment (AIAA) methods cannot effectively extract visual attributes from art images since the existing artistic image datasets don’t provide any information about visual attributes. The lack of visual attributes reduces the interpretability of AIAA methods and limits their performance. To address these problems, a novel artistic image aesthetics assessment assisted by photographic visual attributes is proposed. The proposed method consists of a feature extraction module and a joint prediction module. The feature extraction module pre-trained on a photographic dataset and an artistic image dataset can learn the information of photographic attributes and the generic artistic aesthetic information. The joint prediction module uses a non-local self-attention block to fuse the photographic visual attribute features with general artistic aesthetic features. The fused features are fed into an FC layer for calculating the artistic image aesthetic score. Experimental results indicate that our proposed method outperforms some state-of-the-art AIAA methods.

Self-supervised video hashing aims at generating hash codes and performing fast video content retrieval by leveraging the visual content information inherent in the videos themselves. Most existing methods often overlook the structure-preserving information within the visual content of the videos and thus cannot learn an effective discriminative video representation. In this paper, a Structure-Preserving Video Hashing (SPVH) via a self-supervised Transformer for retrieval is proposed by exploring the global relationships, local relationships, and inter-video relationships in the visual content of videos. In the proposed SPVH, a Transformer-based autoencoder model is used to extract the deep features of the videos. Moreover, a new structure-preserving loss function with the clustering loss, constraint loss, contrastive loss, and reconstruction loss is designed to capture the structural information of the videos. Extensive experiments are conducted on two large-scale video datasets. The results demonstrate the superior performance of our SPVH compared to some state-of-the-art methods.

In recent years, with the access of high-speed broadband, the number of videos on online sharing platforms has increased sharply, and the copyright protection of videos has been widely concerned. However, existing block-based watermarking methods have high sensitivity under geometric deformation and cannot accurately locate the watermark position, which leads to the failure of watermark extraction. Therefore, a robust and blind video watermarking scheme to resist the online sharing platform is proposed. Firstly, DTCWT is used to preprocess the video to reduce the influence of non-geometric signals on the extracted watermark. Secondly, the low frequency domain is divided into rings of different sizes, and the spatial stability region is established to ensure the synchronization of watermark extraction. Finally, the energy of the ring subband is obtained by DCT on the ring, and the watermark embedding is optimized by adjusting the energy relationship of adjacent energy rings to balance the invisibility and robustness of the watermark using the strong correlation of adjacent video frames. The experimental results show that the proposed watermarking scheme can not only resist common video processing operations, but also show good robustness to video online sharing platforms.

Fine-grained image retrieval (FGIR) is an important topic of image retrieval, and its challenge lies in the accurate identification of image objects with minor inter-class differences and considerable intraclass differences. Most existing methods exploit Convolutional Neural Networks (CNNs) to capture fine-grained and coarse-grained information while overlooking the scale variations. To address these issues, a novel method named Hash Generation Network (HGNet) guided by high frequency information is developed to learn crucial details across different scales. The HGNet consists of a High-Frequency Guidance Module (HFGM) and a Hash Generation Module (HGM). The key contribution is the proposed HFGM which integrates the high-frequency information and multi-scale features extracted from the Swin Transformer. As the Swin Transformer can effectively capture global contextual information, its multi-scale features, guided by high-frequency information that contains fine-grained texture details, can represent both fine-grained and coarse-grained details, thereby guiding the HGM in generating discriminative hash codes. Experimental results show that the HGNet outperforms several SOTA FGIR methods in retrieval performance.

Copy detection is crucial for protecting image copyright. This paper proposes a robust image hashing approach via Weighted Saliency Map (WSM) and Laplacian Eigenmaps (LE) (hereafter WSM-LE approach). An important contribution is the WSM construction via the edge map and the saliency map. As the WSM can indicate the interest regions of image, hash calculation based on WSM can provide robustness of our WSM-LE approach. Another contribution is the low-dimensional feature learning by the LE technique. As the LE technique can effectively learn the internal geometric relationships of image, the extracted low-dimensional features can improve discrimination of our WSM-LE approach. In addition, the low-dimensional features are treated as vectors and the vector distances are used to create a compact and encrypted hash. Numerous experiments and comparisons are conducted to confirm the effectiveness and superiority of our WSM-LE approach. The results indicate that our WSM-LE approach has excellent classification and copy detection performances than some baseline approaches.